Go Version 1 now on App Engine
googleappengine.blogspot.com.au
googleappengine.blogspot.com.au
I would have thought that by now AppEngine would have already cycled through all the developers who were ever curious about it and sent them down the Official AppEngine Cycle: Curiosity, Enjoyment, Frustration, Understanding, Enjoyment, Frustration, Disbelief, Anger, Regret, Searching, Toil, then Relief (with the last few steps being part of the migration off of AppEngine and finally the knowledge that you're finally done with it for good.)
In short, I thought we'd all been burned by it already. I honestly can't imagine where they hope to find the next crop of developers from.
Other than that, the app engine itself is definitely interesting. The idea of not having to manage your own infrastructure is appealing. If they could fix their pricing model and add a few features (better SSL support, ability to open raw outgoing connections, etc.) then they'd have something compelling.
I'll put our own app out there http://www.eggtweeter.com as an example. It's the kind of app that needs exactly what GAE is offering in terms of scalability, requires no maintenance of any kind, we've never had to even think about infrastructure etc.
All that being said, yes there are compelling reasons to use another platform for different kinds of apps. We have some ideas for another web app that would be incredibly difficult on GAE and would probably fit in AWS or Heroku's stack better. But for what we need for eggtweeter, GAE seems to actually be a superior choice.
My app, http://sharedcount.com, serves ~100 million API calls a month off of AppEngine, for ~$40/month (or the equivalent of a mid-size MediaTemple virtual server).
You can easily find this information if you search on the gonuts mailing list.
Also learning to is much simpler than learning JavaScript, there are fewer pitfalls and reading the spec is easy.
Gorilla[1] provides nice libraries for web development in Go.